What is tiresome is how sincerely these people insist on being able to make everyone act according to their will, while simultaneously displaying weakness, incompetence, and extreme pettiness. Trying to threaten people into respecting them. The lack of class is just so unsightly.

There was a time when this kind of thing would fly. When the one in charge is a giant orange child-man who can't keep a consistent thought across a single sentence, it makes it clear that the whole thing is narcissistic theatre. It doesn't surprise me that his underlings would try to emulate it, and do a bad job in the process.

I don't like being a part of the reactionary 'orange man bad' crew, but this is really shockingly bizarre. It's not the kind of behaviour you expect from a real leader of a real superpower. And it does make you think - perhaps there's something to be said about the USA not being nearly the power that it once was, and maybe this is what it looks like after you crest the apex of power.

It’s still much preferable to them being strong and competent in subduing everyone to their will.

That comes later.

It’s more like the abusive parents telling the child that they’ll sell him to the scary man at the bus stop every time they want to coerce the child into doing what they want.

Eventually the child develops disrespect for authority.

It’s just so POWERFUL and DANGEROUS that its very aura disrupts the weaker models.

Be careful, MYTHOS might be reading your comment directly from their janitor's iPhone RAM right now. You do not want to upset such a powerful entity!

Oh no, it will BREAK OUT and send me an email when I’m in the park.

How could it be that a political system which remained largely structurally unchanged since the freaking 18th century isn't equipped to deal with everything that has been going on in the world since then?

It must be due to bad actors.

Or a manager will not be needed anymore with no reports to manage.

It’s the crudeness of available management methods at play. Quite exposing for the profession, really (remember lines of code as measure of productivity?).

I recommend reading about the concept of alienation in Marx and Marcuse. They argue that it’s systematically produced, which I can’t prove, but just the observation that it exists, is very insightful.

You can just compile any tail recursive function to a function with a loop and no recursion.

This is in fact how Elm does it! Tail call recursion compiles to a while loop.

That does not address the use case where I find tail recursion most tempting. That would be mutually recursive functions.

If the function can be written as an idiomatic loop I probably would do so in the first place.

You _can_ do trampolines, but that is kind of infectious, or needs to be very explicit with extra code, etc.

Indeed. It's not very efficient though. If I remember correctly Scala does this.

Right but recursion is only a smaller part of why the optimization is important. It means tail-called functions still build on the stack and long function chains—as is common with fp—can overflow

My home pc has 4x8gb ddr4 sticks that I bought years ago. I fully intend to let them sit in there for years to come.

I bought 32 GiB (4x 8 GiB) DDR3 2100 sticks recently for $30 each for a Xeon E3-1275 V2 box.

My main virtualization home lab / vNAS is 512 GiB (16x 32 GiB) DDR4 ECC 3200 I bought 5 years ago. I don't have any need to have or buy more. (Although I had to buy 1 stick last year after a craptastic, used, defective EPYC 7742 burned it out and I went back dual 7402's.)

DDR4lyfe is my new motto.

Fr. I'm selling a matched pair of 48 GiB DDR5 non-ECC 5600 SO-DIMM sticks on secondary markets for $1100. I'm not touching DDR5 again for the foreseeable future, not for 5 years or more. My last foray was 4x 64 GiB 6000 ECC UDIMMs for 2 Ryzen 9 boxes. (Holy shit, that ram is worth $5500 now. It's more than the entire system cost originally including GPUs.)

I'm interested in building a Ryzen 9 box with ECC UDIMMs, but only 2 sticks due to reported stability issues with 4. Did it work well? Would you recommend it? (I would only be running Linux).

This is condescending and wrong at the same time (best combo).

LLMs do stumble into long prediction chains that don’t lead the inference in any useful direction, wasting tokens and compute.

Are you sure about that? Chain of thought does not need to be semantically useful to improve LLM performance. https://arxiv.org/abs/2404.15758

still doesn't mean all tokens are useful. it's the point of benchmarks

If you're misusing LLMs to solve TC^0 problems, which is what the paper is about, then... you also don't need the slop lavine. You can just inject a bunch of filler tokens yourself.